Google I/O Highlights, OpenAI Won’t Stop, Claude in Europe, and Mass Production for Arm AI Chips

PLUS HOT AI Tools & Tutorials

May 17, 2024

Hello and welcome to our weekly digest!

What a week! The confrontation between OpenAI and Google has gone into offline events. Both companies held big keynotes and showed off dozens of interesting products. We realized discussing all topics in one post would be impossible, so we dedicated a separate newsletter to the OpenAI event. You can find it here:

GPT-4o Is Out, And It's Alive?!

Anvar Abrarov

May 14

Read full story

So today, we’ll talk about everything Google pleased us with during I/O. And, of course, we won’t forget about other important news of this week. Let's get started!

This Creators’ AI Edition:

Featured Materials 🎟️
News of the week 🌍
Useful tools ⚒️
Weekly Guides 📕
AI Meme of the Week 🤡
AI Tweet of the Week 🐦
(Bonus) Materials 🎁

Featured Material 🎟️

Google I/O Highlights

Yep, just one question: where do we start? Google has had the most eventful I/O keynote in its history. The company talked a lot about AI (and I mean a lot), showed off some cool new platforms, and spoke about AI updates for available products. To give you an insight, Google mentioned “AI” 121 times. All in 110 minutes.

It's pretty funny that the number of mentions was counted by **Google,** using AI.

Let's try to discuss all the major announcements of this event.

New Gemini Models

Google's first big announcement is the new Gemini 1.5 Flash model. As you can guess by the name, this is the company's fastest model. That said, Google says the Flash is as powerful as the Gemini Pro. It's designed for “narrow, high-frequency, low-latency tasks,” while the Gemini 1.5 Pro model suits activities without quick answers.

As The Verge rightly pointed out, Gemini 1.5 Flash might be a better option for real-time responses to customers or fast image generation, while Gemini 1.5 Pro could read and summarize research papers.

The Gemini 1.5 Pro itself has also received some major improvements. In addition to expanding the context window to 2 million tokens, Google has improved code generation, logical reasoning, multi-way dialog, and audio and image understanding.

The company also added audio understanding to Gemini API and AI Studio, so 1.5 Pro can now analyze images and audio for videos uploaded to AI Studio. Google is integrating 1.5 Pro into other products, including Gemini Advanced and Workspace.

Both models will soon be available in over 200 countries.

It's pretty funny that Google 1.5 Flash came out a day after the GPT-4o was announced. Now these models look like direct competitors.

Keep your mailbox updated with key knowledge & news from the AI industry

AI in Google Search

Another major announcement that affects us all is the search update. Google is starting to roll out “AI Overviews”. Previously, you may have heard of this platform as the Search Generative Experience. Google's search engine will soon start generating AI summaries at the top of the search page.

And that's just a small part of what the future holds.

AI Overviews in Search are rolling out to everyone in the U.S. beginning this week, and more countries will follow soon.

Google Lens has also gotten noticeably more interesting. Users now search the web by uploading a video through the app. All you have to do is turn on the recording, point the camera at the desired object, and explain what you want to find.

https://duet-cdn.vox-cdn.com/thumbor/0x0:3000x2000/1200x800/filters:focal(1500x1000:1501x1001):format(webp)/cdn.vox-cdn.com/uploads/chorus_asset/file/25447243/Google_Lens_video_search.jpg

As an example of a useful case study, Google tells that you can use Lens to determine the cause of a car breakdown or find a product you want to buy and order it in a couple of seconds.

Google Veo (OpenAI Sora' Competitor)

Veo is Google's most powerful video generation model. It generates 1080p clips in a wide range of visual styles and over one minute in length. The AI understands cinematic terms like "timelapse" or "aerial landscape filming." It creates consistent shots so people, animals, and objects move realistically throughout the frame.

Here's a preview of Google's work with filmmaker Donald Glover and his creative studio, Gilga, who experimented with Veo for a movie project.

It didn't take very long for Sora to be the one and only video generation solution. And that's great - with serious competition, progress will be even faster.

Project Astra

Project Astra is a visual chatbot that Google calls "the future of AI assistants". And we can see why. This experimental platform combines the new Lens features with a real-time chatbot. The AI analyzes the environment around the user and answers questions related to objects on the screen.

It's not just about describing objects (we've seen this before and more than once). Astra can use Google Maps to determine the user's location, accurately describe all objects in a room, and, for example, explain the code the person writes.

Here's how it works:

I'm eager to try a lot of the products presented at I/O, but Project Astra looks like the most impressive thing. The first thing that comes to mind is how cool it would work in combination with smart glasses or lenses.

What else?

Apart from the updates I've described above:

Google has announced Trillium. It’s the sixth generation of a custom AI accelerator, the Tensor Processing Unit (TPU);
Grounding with Google Search is now generally available on Vertex AI;
New high-quality image generation model Imagen 3;
VideoFX — a new experimental tool that uses Google DeepMind’s generative video model to turn an idea into a video clip;

And much more. If you can't get enough, Google's blog has a separate page called "100 things we announced at I/O 2024". You'll find all the updates here.

And you can also watch a recap made by Google:

Sharing is caring! Refer someone who recently started a learning journey in AI. Make them more productive and earn rewards!

Refer a friend

News Of The Week 🌍

Bloomberg: OpenAI Finally Makes a Deal with Apple

We've already covered how Apple talked with OpenAI to integrate ChatGPT into iOS 18. But back then, those were just rumors of negotiations. Well, now it's rumors of a deal being done. According to Bloomberg, the companies have come to an agreement, and Apple will soon announce chatbot support for the iPhone.

We'll learn more about this update at WWDC 2024 this summer.

Bloomberg also says that Apple has also held talks with Google about licensing the Gemini chatbot. Those discussions have not led to an agreement, but are ongoing, anonymous insiders told us.

OpenAI Inks Deal to Train AI on Reddit Data

Like I said in the title, OpenAI won't stop. So, let's get to the next big deal involving our favorite startup. After many rumors and leaks, we got confirmation that OpenAI will be using Reddit data to train new models.

In the near future, OpenAI will also bring Reddit's content to ChatGPT and other products. At the same time, some AI features will appear on Reddit that should improve users' experience. OpenAI will also become a Reddit advertising partner.

No other details are available yet, but the heads of OpenAI and Reddit promise to share them soon.

Well, now OpenAI has gotten its hands on one of the largest datasets on the internet. Let's hope the partnership with Reddit will make GPT more useful.

Cofounder Ilya Sutskever Leaves OpenAI

It's a big loss for OpenAI. Ilya Sutskever, one of the founders and a leading researcher, has announced that he has decided to leave the company. He didn’t disclose the reason but cited that he would develop his own project. Sam Altman expressed regret about Sutskever's departure from the company, calling him not only one of the outstanding minds of our time but also a close friend.

"I am excited for what comes next - a project that is very personally meaningful to me about which I will share details in due time," Sutskever wrote in an X post.

Jakub Pachocki will now fill Sutskever's position.

In case you've forgotten, last year Sutskever disagreed with Sam Altman's policies, and initially co-sponsored his resignation, but then sided with those who insisted on his return. I don't want to speculate, but it looks like these events are related.

SoftBank's Arm plans to launch AI chips in 2025

Big news from the tech manufacturing industry. According to Nikkei Asia, Arm will create a division to produce AI chips. The first prototypes will be ready by spring 2025. TSMC (the same one that assembles the best chips for Apple, Google, and many others) will be responsible for mass-producing these processors. Arm clients will start getting their chips next fall.

Softbank (Arm's parent company) is rumored to be very serious about AI. CEO Masayoshi Son already intends to invest $64 billion in software, data centers, and robotics to support the development of advanced algorithms. These products promise to affect both technology companies and ordinary users.

Creators’ AI Subscription is the best gift for Creators diving into AI.

Give a gift subscription

ElevenLabs Launches AI-Voiced Screen Reader App

Unicorn startup ElevenLabs has quietly released its first consumer app. Of course, it's related to machine learning and voice generation. ElevenLabs Reader: AI Audio allows users to recognize and voice text from web pages, PDFs, and other documents. The app is already available for download on the App Store.

It supports 11 voice options.

Apparently, Eleven Labs is ready to go beyond the "famous voice from TikTok" and offer simple but useful tools for regular users. Sounds nice.

Anthropic Launches Claude AI Across Europe

The company has launched a global expansion. Anthropic says European individuals and businesses can now access its Claude chatbot through the web and an iOS mobile app. The company has also launched business-oriented Claude Team subscription plans, priced at 28 euros ($30). Claude's main features for business are accuracy, security, and privacy.

Anthropic separately noted that Claude is now fluent in French, German, Italian, Spanish, and other European languages.

Useful Tools ⚒️

Glato AI – Create winning video ads with AI creators from a product URL

Oliv – Save 10 hours/week in tiring sales tasks & Gong call reviews

AI Link Manager – Detect, fix, and deploy internal links in a click

FISKL – AI-powered accounting & financial management

Voicenotes – truly intelligent AI note-taker.

God knows how many note-taking apps I've tried, and I didn't like each one for some reason. So maybe I was looking in the wrong direction? Voicenotes is a new AI platform for ideas, tasks, meetings, etc. This chatbot helps you create and structure any important information. Your notes are automatically titled, neatly organized, and synchronized across devices.

I'm already trying it out regularly and highly recommend you do the same.